Package com.datarobot.prediction.internal

Class TimeSeriesRegressionPredictorImpl

java.lang.Object

com.datarobot.prediction.internal.TimeSeriesBase

com.datarobot.prediction.internal.TimeSeriesRegressionPredictorImpl

All Implemented Interfaces:

IEventProvider<TimeSeriesRegressionScoreEventInfo>, IPredictorInfo, ITimeSeriesModelInfo, ITimeSeriesRegressionPredictor, Serializable

public class TimeSeriesRegressionPredictorImpl
extends TimeSeriesBase
implements ITimeSeriesRegressionPredictor

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
protected LinkedHashMap<String,Class>	features
protected Map<String,Integer>	featuresIndexes
protected String	modelId
protected Map<String,String>	modelInfo
protected int	numDoubles
protected int	numStrings
protected ConcurrentHashMap<String,String>	sanitizedFeaturesCache
protected com.datarobot.prediction.internal.PredictorBase.GenericEvent<ScoreEventT>	scoreEvent
protected String	targetName

Fields inherited from class com.datarobot.prediction.internal.TimeSeriesBase

dateColumnName, dateFormat, fdw, fw, seriesIdColumnName, timeStep

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	TimeSeriesRegressionPredictorImpl(TimeSeriesPredictor predictor)

Method Summary

Modifier and Type	Method	Description
protected Row	extractFeatures(Iterable<?> row)
protected Row	extractFeatures(Map<String,?> row)	Extracts features that are needed from initial row.
protected Map<String,?>	extractFeaturesMap(Iterable<?> row)
String	getDateColumnName()	Getter for Date Column Name.
String	getDateFormat()	Expected date format for the input dates.
protected String	getDefaultEmptyColumnName()	Return label for empty column.
protected ExplanationParams	getDefaultParams()
ExplanationParams	getDefaultPredictionExplanationParams()	Retrieves default prediction explanation parameters.
protected Set<String>	getDoubleMissingValuesLabels()	Returns the set of null labels.
protected Double	getDoubleValue(String featureName, Object value)	Replaces double missing value with NaN.
Map.Entry<Integer,Integer>	getFeatureDerivationWindow()	Getter for Feature Derivation Interval (FDW).
LinkedHashMap<String,Class>	getFeatures()	Model specific features that are required for scoring.
protected Field	getField(Class<?> clazz, String fieldName)
Map.Entry<Integer,Integer>	getForecastWindow()	Getter for Forecast Interval.
String	getModelId()	DataRobot Model ID.
Map<String,String>	getModelInfo()	Map with different model information.
Class<?>	getPredictorClass()	What the type of the model
protected Method	getPredictorMethod(String methodName, Class<?>... signature)
Event<TimeSeriesRegressionScoreEventInfo>	getScoreEvent()	Returns the event that is invoked for each of the score calculated by the model.
String	getSeriesIdColumnName()	Getter for Series ID Column Name.
Map.Entry<Integer,String>	getTimeStep()	Getter for Time Step.
protected Map<String,String>	initializeModelInfo(PredictorT predictor)
protected String	sanitize(String name)	Some pre-processing for feature name.
List<TimeSeriesScore<Double>>	score(ArrayList<Map<String,?>> rows)	Calculates score for the Time-Series rows.
List<TimeSeriesScore<Double>>	score(ArrayList<Map<String,?>> rows, TimeSeriesOptions options)	Generalized method for calculating predictons could be used for any type of predictions.
List<TimeSeriesScore<Double>>	score(ArrayList<Map<String,?>> rows, String forecastPoint)	Calculates score for the Time-Series rows.
List<TimeSeriesScore<Double>>	score(ArrayList<Map<String,?>> rows, String startDate, String endDate)	Calculates historical scores for the Time-Series rows.
protected void	sendScoreInfo(ScoreEventT eventInfo)

Methods inherited from class com.datarobot.prediction.internal.TimeSeriesBase

extractRows

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

numDoubles

protected int numDoubles

numStrings

protected int numStrings

features

protected LinkedHashMap<String,Class> features

featuresIndexes

protected Map<String,Integer> featuresIndexes

sanitizedFeaturesCache

protected ConcurrentHashMap<String,String> sanitizedFeaturesCache

modelInfo

protected Map<String,String> modelInfo

modelId

protected String modelId

targetName

protected String targetName

scoreEvent

protected final com.datarobot.prediction.internal.PredictorBase.GenericEvent<ScoreEventT extends EventInfo> scoreEvent

Constructor Detail

TimeSeriesRegressionPredictorImpl

protected TimeSeriesRegressionPredictorImpl(TimeSeriesPredictor predictor)

Method Detail

score

public List<TimeSeriesScore<Double>> score(ArrayList<Map<String,?>> rows)

Description copied from interface: ITimeSeriesRegressionPredictor

Calculates score for the Time-Series rows. NOTE: rows will be sorted lexicographical order by series_id then by date so adds linear overhead if all data is already sorted and historic data is extracted from the rows to predict by the setted/unsetted to not NaN target values.

Specified by:

score in interface ITimeSeriesRegressionPredictor

Parameters:

rows - time-series that contains both historic data and points to predict.

Returns:

collection of scores only for the rows to predict.

score

public List<TimeSeriesScore<Double>> score(ArrayList<Map<String,?>> rows,
                                           String forecastPoint)

Description copied from interface: ITimeSeriesRegressionPredictor

Calculates score for the Time-Series rows. NOTE: rows will be sorted lexicographical order by series_id then by date so adds linear overhead if all data is already sorted. Historic data is extracted from the rows to predict by the forecastPoint.

Specified by:

score in interface ITimeSeriesRegressionPredictor

Parameters:

rows - time-series that contains both historic data and points to predict.

forecastPoint - date from which to start predictions.

Returns:

collection of scores only for the rows to predict.

score

public List<TimeSeriesScore<Double>> score(ArrayList<Map<String,?>> rows,
                                           String startDate,
                                           String endDate)

Description copied from interface: ITimeSeriesRegressionPredictor

Calculates historical scores for the Time-Series rows. NOTE: rows is expected to be sorted by the date, and contains only historic data.

Specified by:

score in interface ITimeSeriesRegressionPredictor

Parameters:

rows - history time-series.

startDate - start date for historical predictions.

endDate - end date for historical predictions.

Returns:

collection of scores only for the rows between startDate and endDate.

score

public List<TimeSeriesScore<Double>> score(ArrayList<Map<String,?>> rows,
                                           TimeSeriesOptions options)

Description copied from interface: ITimeSeriesRegressionPredictor

Generalized method for calculating predictons could be used for any type of predictions.

Specified by:

score in interface ITimeSeriesRegressionPredictor

Parameters:

rows - - time series to use for calculate predictions.

options - - ts scoring options

Returns:

predictions according to the options provided.

getSeriesIdColumnName

public String getSeriesIdColumnName()

Description copied from interface: ITimeSeriesModelInfo

Getter for Series ID Column Name.

Specified by:

getSeriesIdColumnName in interface ITimeSeriesModelInfo

Returns:

series id used during training.

getDateColumnName

public String getDateColumnName()

Description copied from interface: ITimeSeriesModelInfo

Getter for Date Column Name.

Specified by:

getDateColumnName in interface ITimeSeriesModelInfo

Returns:

date column name used for indexing.

getTimeStep

public Map.Entry<Integer,String> getTimeStep()

Description copied from interface: ITimeSeriesModelInfo

Getter for Time Step. Pair of quantity to time unit. Time Unit could be: - ROW - MILLISECOND - SECOND - MINUTE - HOUR - DAY - WEEK - MONTH - QUARTER - YEAR

Specified by:

getTimeStep in interface ITimeSeriesModelInfo

Returns:

time step used for the time series processing.

getFeatureDerivationWindow

public Map.Entry<Integer,Integer> getFeatureDerivationWindow()

Description copied from interface: ITimeSeriesModelInfo

Getter for Feature Derivation Interval (FDW). See documentation https://app.datarobot.com/docs/time/time-series.html#framework Pair of min units number and max units number. So window is calculated date based formulas similar to: [ forecast_point - max * time_step_millis, forecast_point - min * time_step_millis ]

Specified by:

getFeatureDerivationWindow in interface ITimeSeriesModelInfo

Returns:

feature derivation window used for feature extraction.

getForecastWindow

public Map.Entry<Integer,Integer> getForecastWindow()

Description copied from interface: ITimeSeriesModelInfo

Getter for Forecast Interval. Pair of min units number and max units number So window is calculated date based formulas similar to: [ forecast_point + min * time_step_millis, forecast_point + max * time_step_millis ]

Specified by:

getForecastWindow in interface ITimeSeriesModelInfo

Returns:

feature derivation window used for feature extraction.

getDateFormat

public String getDateFormat()

Description copied from interface: ITimeSeriesModelInfo

Expected date format for the input dates.

Specified by:

getDateFormat in interface ITimeSeriesModelInfo

Returns:

date format.

getModelId

public String getModelId()

Description copied from interface: IPredictorInfo

DataRobot Model ID.

Specified by:

getModelId in interface IPredictorInfo

Returns:

string model id.

getFeatures

public LinkedHashMap<String,Class> getFeatures()

Description copied from interface: IPredictorInfo

Model specific features that are required for scoring. The order of entries is important.

Specified by:

getFeatures in interface IPredictorInfo

Returns:

Map of feature names and expected type in the order.

getPredictorClass

public Class<?> getPredictorClass()

Description copied from interface: IPredictorInfo

What the type of the model

Specified by:

getPredictorClass in interface IPredictorInfo

Returns:

interface class that implements that particular model.

getDefaultPredictionExplanationParams

public ExplanationParams getDefaultPredictionExplanationParams()

Description copied from interface: IPredictorInfo

Retrieves default prediction explanation parameters.

Specified by:

getDefaultPredictionExplanationParams in interface IPredictorInfo

Returns:

Default prediction explanation params, null if no defaults available.

getModelInfo

public Map<String,String> getModelInfo()

Description copied from interface: IPredictorInfo

Map with different model information.

Specified by:

getModelInfo in interface IPredictorInfo

Returns:

Map with different model information

getScoreEvent

public Event<TimeSeriesRegressionScoreEventInfo> getScoreEvent()

Description copied from interface: IEventProvider

Returns the event that is invoked for each of the score calculated by the model.

Specified by:

getScoreEvent in interface IEventProvider<TimeSeriesRegressionScoreEventInfo>

Returns:

scoring event information.

extractFeatures

protected Row extractFeatures(Map<String,?> row)

Extracts features that are needed from initial row.

Parameters:

row - - Map of initial feature name to its value.

Returns:

- internal type for scoring.

extractFeatures

protected Row extractFeatures(Iterable<?> row)

extractFeaturesMap

protected Map<String,?> extractFeaturesMap(Iterable<?> row)

sanitize

protected String sanitize(String name)

Some pre-processing for feature name.

Parameters:

name - - original feature name.

Returns:

- sanitized name.

getDoubleValue

protected Double getDoubleValue(String featureName,
                                Object value)

Replaces double missing value with NaN.

Parameters:

featureName - - name of the feature.

value - - object feature value.

Returns:

- parsed Double value.

getDefaultEmptyColumnName

protected String getDefaultEmptyColumnName()

Return label for empty column.

Returns:

- empty column name.

getDoubleMissingValuesLabels

protected Set<String> getDoubleMissingValuesLabels()

Returns the set of null labels.

Returns:

- set of null labels.

getField

protected Field getField(Class<?> clazz,
                         String fieldName)

initializeModelInfo

protected Map<String,String> initializeModelInfo(PredictorT predictor)

getPredictorMethod

protected Method getPredictorMethod(String methodName,
                                    Class<?>... signature)

sendScoreInfo

protected void sendScoreInfo(ScoreEventT eventInfo)

getDefaultParams

protected ExplanationParams getDefaultParams()